batch_size=4
num_epochs = 10
warmup_steps = 1
loss: MultipleNegativesRankingLoss
base_model:TownsWu/PEG
